Counter argument: Not only is taste subjective, it's a moving target. As soon as [tasteful thing] becomes easily replicable, it stops being tasteful. Taste is a status game with a strong defensibility mechanism. That's why the term is so fuzzy and ill-defined. Vagueness is the whole point. It's meant to be hard to copy.

ridiculously good times with friends from taxa tech, merge labs, skild robotics, and more. ofc special thanks to @SumnerLN for making me some rolls while i snapped pics 🍣 📸

we just shipped a new look for @fearnlegal ! we rebuilt the aesthetic of every screen around one idea: that the software should look and feel like a reliable and precise technical instrument. amazing work by @kvnbnntt and @broooyannn!

genuinely there are so many interesting problems in the ai infra world. it just so happens that now inference is trendy. but there are a ton of other ones. you aren’t bullish enough of infrastructure jobs.

Either the jev model is miscalibrated here, or I am thinking about this wrong. @typesafeai, should the result here be 50/50, or is my understanding of what the model should output wrong?

The days of waiting weeks for a patent are over. Today we’re launching Fearn, a modern patent firm with flat fees, turnaround in days, and human expertise embedded in every part of the application cycle, all built on our own operating system, FearnOS. One detail I'm especially proud of: in our editor, expert edits can't be overwritten by AI. Technology executes the mechanical work while people make the judgment calls. That's how we move fast without cutting corners. Legal services are at an inflection point. We think founders deserve a new standard, and we're excited to start delivering it.

Introducing pgrust v0.3: the bug-smashing release. We evaluated pgrust v0.2 and v0.3 against the same held-out corpus: 164,000 test cases from 50 open-source projects. pgrust v0.2 diverged from Postgres on 1,454 cases. pgrust v0.3 diverged on 11. pgrust is 300x faster than Postgres for analytical workloads, and we're a drop-in replacement. Held-out corpus included: @djangoproject, SQLAlchemy, Hibernate, @rails, and more. @pgrustdb is now much more Postgres-compatible and reliable than before. @JasonSeibel and I don't recommend running pgrust in production (yet), but we are successfully running it ourselves internally. - @AntithesisHQ and @workersdotio for simulation testing - Detail.dev by @danlovesproofs for LLM-assisted code analysis - Multiple home-built differential fuzzers - Kani for formal verification Our goal is for pgrust to be truly 100% compatible with Postgres. There are still bugs. You should still run Postgres for any data you can't afford to lose. But we want people trying pgrust on real workloads in non-critical environments. If you need Clickhouse performance queries in Postgres, are frustrated by the many gotchas in Postgres, or just geek out about databases and Rust, feel free to send me a DM as we would love to chat.

dinner with hardware founders last night courtesy of @fearnlegal. no panels, no pitches, just people who actually build things for the love of the game. more to come!

another great article by @SamRedfordFearn!

lucky to be working with a couple breakthrough-level life science day 0 startups that are looking for AI researcher talent who want to work on something meaningful. DMs open for that if you have recommendations, or want to hear more.

i have mixed thoughts on the navier stokes situation. on the one hand, it's really exciting to see massive progress on such a famous PDE, but the context of how it came about makes me really sad about what it means to do science in this era as someone trained in academia, it feels like an existential threat for AI labs to be interested in your line of work because the novelty of your ideas is how you are evaluated. unlike in industry, being the first to make a discovery is how credit is assigned. it’s a shame to see underhanded behavior in order to be the first to make the discovery regardless of my personal feelings, protecting innovation and open science is critical now more than ever. the reason why many smart people choose to go to and stay in academia is because of open science, and as a society, we reap the benefits of open science one of the best ways to continue to do open science while protecting yourself is through IP or through open source AI providers, especially when there is no way to guarantee that your IP is safe from frontier labs i wonder in the long run, will science become more closed or will there be an influx of unvetted results just to be the one to plant the flag?

we're rolling into tech week with the coolest experience 🍣 see ya on 10/8 partiful.com/e/Wn9UsaQDG7NeK…

interesting how this post translates to vertical AI services businesses. in the short term, it's sufficient for the service provider (patent agents in our case) to do the service work as they have always done. ultimately, this is suboptimal in the long term as competitors become AI-native. thus, it is a necessary requirement to build repeatable AI work flows that work, and it's the job of the engineer who owns a feature to figure out how to do just that

we've believed in this since we started. of course, this is only enabled either by 1. scale to find the subproblems or 2. understanding the subproblems from the get go. faster, cheaper, and more reliable

such a wonderful evening with founders across various scientific domains! the one thing that we agreed on is that AI for science is not solved yet :) it felt like being back at grad school talking to peers who work on hard problems that I didn't know anything about. if this is something that you enjoy, join us for the next one

people often ask me if I would recommend that they do a phd now. even when i started in 2019, I wouldn't have recommended doing a phd in CS because of the opportunity cost, and now it's even harder to justify for most people. however, doing a phd is such a unique opportunity to work on hard problems purely because they are interesting, and you learn a lot about yourself in doing so. looking back, my phd was both extremely challenging but also deeply rewarding, and i'm so grateful to my amazing phd advisor @klbouman for giving me so much intellectual freedom. now more than ever, it is important to work on hard problems with unclear answers.